Detecting Attacks in V2G Environments Using an Isolation Forest Based Anomaly Detection Model

In this paper, we investigate communication security in the Vehicle to Grid (V2G) network. We assume connection of electric vehicles with a charging network via aggregators. In particular, we propose isolation forests as an anomaly detection model based on unsupervised learning. This model detects anomalies in cyber messages corresponding to various state changes and data constraints, with low linear time complexity and low memory requirements. We trained the proposed model and evaluated its effectiveness using a V2G database obtained from the MiniV2G project. The experimental results revealed that the proposed model is effective in terms of detection score, precision, recall (detection rate), and false positive rate (false alarm rate) on this specific dataset. Compared to two other anomaly detection algorithm our model gave better results.